Lung cancer: an AI reads the scanners and guesses 26 months in advance the gain in survival of a treatment

In the phase III MARIPOSA trial, Altis Labs' IPRO tool detected a benefit from the sixteenth week that conventional imaging criteria had not seen come.

Altis Labs presented to the World Congress on Lung Cancer (WCLC) in Seoul the results of a retrospective analysis of the phase III MARIPOSA trial on non-small cell bronchial cancer with EGFR mutation. His IPRO tool automatically analyses the scanners already performed in the test — about 10,000 tests here — and draws a survival prediction, where the usual RECIST criteria simply measure the size of the lesions.

The signal detected in the 16th week by the IPRO response rate announced the final benefit of treatment: 11 months before progression-free survival data were available, and 26 months before the maturation of the overall survival data. The conventional objective response rate had not anticipated this gain, while the median survival gap would eventually exceed 12 months.

For Felix Baldauf-Lenschen, founder and CEO of Altis Labs, the demonstration focuses less on diagnosis than on conducting tests: an automated imaging marker able to cut sooner would reorient or stop a trial without waiting for years. Non-small cell lung cancer accounts for 85% of lung cancers, and five-year survival remains below 20% in advanced forms mutated EGFR despite targeted therapies.